Accuracy of Dem-Based Topographic Data in Flood Inundation Modelling: A Case of Wami River, Tanzania

The objective of this paper is to evaluate and further comment on the accuracy of the DEM-based topographic data as input in flood modelling. Recent studies have indicated thathydraulic parameters of the modified cross sections based on high resolution DEM aregeometrically and hydraulically similar to the measured ones. In the current study hydraulicmodelling was intended to guide a development project whereby a floodplain protectionembankment is contemplated. Therefore, accuracy in terms of positioning/alignment of naturalfeatures such as river banks and heights/elevations was imperative to attain. In order to achievethese, accuracy of various topographic data sources were qualitatively and/or quantitativelyanalysed before and after hydraulic modelling. As the study river reach is ungauged, thehydraulic model was verified and validated with field observations, readily availableinformation and professional judgement. Based on error analysis, it was found thatuncalibrated 90-m resolution STRM DEM-based topographic data are biased, i.e., 80% of theerror mass is negative. The error increases with elevation. The final input to the hydraulicmodel was a composite/hybrid DEM derived after merging geometric data from fourrepresentative field-measured cross sections, a 2-m interval contour map covering part of theriver valley, and calibrated 90-m resolution DEM. Although, calibration improves the data, itdoes not remove completely the inherent bias. Further to DEM calibration, it was thereforerecommended to include an elevation uncertainty value of -0.5 m, as derived from topographicanalysis, into simulated water surface profiles elevations. The hydraulic model satisfactorilysimulated flood inundation extent after considering the elevation uncertainty. In order toimprove the findings further studies should use adequate measured cross sections and DEM ofhigher resolutions.